Question

namespace A
{
    class B
    {
    public:
        B(int);
    };
}

namespace C
{
    class D
    {
        static const ::A::B b;
    };
}

const ::A::B ::C::D::b(0);      // #1
// const ::A::B C::D::b(0);     // #2
// const ::A::B (::C::D::b)(0); // #3

Option #1 fails to compile. Now that I’ve thought about it for a while, I’m pretty sure the compiler considers the whitespace between “B” and “::C” insignificant, so it is trying to find a member “C” inside “B”. If that’s what’s happening, I need some way to convince the compiler that these are two separate qualified names.

Options #2 and #3 both seem to work in a very minimal test. #3 seems somewhat less vulnerable to name collisions since it’s more qualified. #3 is also somewhat easier for me to switch to. So I lean towards #3, but it looks weird.

Is there any strong reason to prefer one over the other? Can I expect both to work correctly on other compilers? Is there an even better solution than either one? And for that matter, am I correct about why #1 fails?

Probably unnecessary details

• The code that inspired this question was output by a source code generator I wrote. The identifiers are derived from the generator’s input, so I’m concerned about name collisions, specifically unintentional shadowing between scopes, which the generator can’t detect. So, I wrote the generator to qualify every identifier as fully as possible.
• I’m compiling on VC2010.
• I’m intentionally not using C++0X features, as I want the code to be portable to certain older compilers.
• The specific compiler error is “error C3083: ‘C’: the symbol to the left of a ‘::’ must be a type”

Answer 1

I’m pretty sure the compiler considers the whitespace between “B” and “::C” insignificant, so it is trying to find a member “C” inside “B”

That’s precisely correct.

Is there any strong reason to prefer one over the other? Can I expect both to work correctly on other compilers?

Option #3 looks fine to me (and is compliant); that is, if you’re really stuck with this rather silly-looking requirement. 🙂

I use it myself to conclude a blog post on a very similar “issue”.